id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
464b3da8-ede1-4723-a0e3-f1c37b5336bd | trentmkelly/LessWrong-43k | LessWrong | 20000 UC Berkeley lectures that were deleted from Youtube
|
1f466f55-bb42-4e01-89c2-0a6dce71d8b3 | trentmkelly/LessWrong-43k | LessWrong | The Structure of the Pain of Change
If there's something in your life that isn't quite good enough and you'd like to change it, chances are you can't just swap one thing for another directly. There are many reasons for that, but this post won’t focus on the various obstacles or how to overcome them. Instead, we'll try... |
e58d2688-eabb-4f14-9edd-89ff343ca0b5 | trentmkelly/LessWrong-43k | LessWrong | Toy Models and Tegum Products
(Thanks to Adam Scherlis, Kshitij Sachan, Buck Shlegeris, Chris Olah, and Nicholas Schiefer for conversations that informed this post. Thanks to Aryan Bhatt for catching an error in the loss minimization.)
Anthropic recently published a paper on toy models of superposition [Elhage+22]. O... |
3cc4c70c-46c4-4826-bf14-ac978a466c78 | trentmkelly/LessWrong-43k | LessWrong | Review of AI Alignment Progress
I'm having trouble keeping track of everything I've learned about AI and AI alignment in the past year or so. I'm writing this post in part to organize my thoughts, and to a lesser extent I'm hoping for feedback about what important new developments I've been neglecting. I'm sure that I... |
d545acfe-affa-4a29-9752-b76953eb1790 | trentmkelly/LessWrong-43k | LessWrong | Long-form data bottlenecks might stall AI progress for years
Cross-posted from my Substack.
Introduction
AI 2027’s recent publication made waves—if you’re reading this, you probably saw the whole saga unfold: cheers, jeers, NYT writeup, the works. To briefly recap: the authors predict that by March 2027, a superhu... |
4ee9dbcf-54fa-4a64-81e4-6dc72f06e2fd | trentmkelly/LessWrong-43k | LessWrong | AI safety logo design contest, due mid-May
Please have a look at the updated version of this post on the EA Forum.
This contest is about creating a logo to represent AI safety research.
I’m being supported by an independent funder and people from nonlinear.org and am taking care of administration and communication.... |
c2da4bea-17e1-4a01-a2b1-6ad6d91fbb9c | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Ban development of unpredictable powerful models?
In this post, I propose a (relatively strict) prediction-based eval which is well-defined, which seems to rule out lots of accident risk, and which seems to partially align industry incentives with real alignment progress. This idea is not ready to be implemented, but ... |
180ecd43-a66b-4354-a172-b5aa82ecd382 | StampyAI/alignment-research-dataset/special_docs | Other | Why Does Deep and Cheap Learning Work So Well?
J Stat Phys
DOI 10.1007/s10955-017-1836-5
Why Does Deep and Cheap Learning Work So Well?
Henry W. Lin1·Max Tegmark2·David Rolnick3
Received: 3 December 2016 / Accepted: 27 June 2017
© Springer Science+Business Media, LLC 2017
Abstract We show how the success of deep learn... |
c151fbf5-71b5-4c33-93b8-e09794184128 | trentmkelly/LessWrong-43k | LessWrong | When do "brains beat brawn" in Chess? An experiment
As a kid, I really enjoyed chess, as did my dad. Naturally, I wanted to play him. The problem was that my dad was extremely good. He was playing local tournaments and could play blindfolded, while I was, well, a child. In a purely skill based game like chess, an extr... |
a4ac8e80-f71f-4fa6-8289-7152a32a08ed | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | SERI ML Alignment Theory Scholars Program 2022
The Stanford Existential Risks Initiative ([SERI](https://cisac.fsi.stanford.edu/stanford-existential-risks-initiative/content/stanford-existential-risks-initiative)) recently opened [applications](https://airtable.com/shr3XO3xZVR9HNPS2) for the second iteration of the [M... |
7258c054-2725-479d-8289-3664fd623828 | trentmkelly/LessWrong-43k | LessWrong | Non-standard cryo ideas
What plans could a prospective cryonicist try out, beyond simply signing up, that could increase the odds of eventually having a pleasant re-animation experience?
To show what I mean, here are the main ideas I've managed to come up with so far. None of these particular ideas are a standard par... |
25cd92b0-e56a-4b80-ac52-83e47b4b4b34 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Madison Monday Meetup
Discussion article for the meetup : Madison Monday Meetup
WHEN: 12 March 2012 05:30:00PM (-0600)
WHERE: 1831 Monroe St., Madison, WI
This time around, let's see if we can find ways to be more "strategic", as in: http://lesswrong.com/lw/2p5/humans_are_not_automatically_strategic/
I'd ... |
a6a84b58-6169-4d8b-97aa-2f941c0d431e | trentmkelly/LessWrong-43k | LessWrong | Communicating via writing vs. in person
There's a lot that I really like about communicating via writing. Communicating in person is sometimes frustrating for me, and communicating via writing addresses a lot of those frustrations:
1) I often want to make a point that depends on the other person knowing X. In person,... |
7c873033-5d40-46da-93d4-7cc8cba0f871 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Formalizing the Problem of Side Effect Regularization
1 Introduction
---------------
We need to build *aligned* ai systems, not just capable ai systems. For example, recommender systems which maximize app usage might provoke addiction in their users. Users need the ai system’s behavior to be aligned with their inte... |
c78d81fe-dc4f-42f7-8a16-903d46f0e1c1 | trentmkelly/LessWrong-43k | LessWrong | How to learn a new area X that you have no idea about.
This guide is in response to a request in the open thread. I would like to improve it; If you have some improvement to contribute I would be delighted to hear it! I hope it helps. It was meant to be a written down form of; "wait-stop-think" before approaching a... |
4a8e960f-5b75-4a0a-90e6-00cef4b7fb50 | trentmkelly/LessWrong-43k | LessWrong | "Where Am I?", by Daniel Dennett
”Where Am I?” is a short story by Daniel Dennett from his book Brainstorms: Philosophical Essays on Mind and Psychology. Some of you might already be familiar with it.
The story is a humorous semi-science fiction one, where Dennett gets a job offer form Pentagon that entails moving h... |
2af4da7f-ea18-4300-b932-117e65f07f42 | trentmkelly/LessWrong-43k | LessWrong | AI development is an act of social revolution
Assume the goal of AI research is to build superintelligent decision-making systems. If we successfully develop and deploy such systems, it will likely no longer make sense for humans to work many, or even most, jobs. The duties humans must still perform will radically di... |
d1e28a98-3d15-45b0-97a6-44847e7ad7bd | trentmkelly/LessWrong-43k | LessWrong | February 2012 Media Thread
There was a recent discussion considering the idea of a monthly Book (later expanded to movies, links, etc) thread. The poll was pretty unanimous (Both before and after I was karmassassinated) that this was A Good Idea (tm), and daenerys's Janurary thread (from which I obviously stole a lot ... |
af9b0782-9218-4899-a41e-638d237aba32 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Quantifying Harm
1 Introduction
---------------
AI systems are playing an ever-expanding role in making decisions, in
applications ranging from hiring and interviewing to healthcare to
autonomous vehicles. Perhaps not surprisingly, this is leading to
increasing scrutiny of the harm and benefit caused by (the decisi... |
6910eb00-3e91-4aa9-ba78-dba22dd31679 | trentmkelly/LessWrong-43k | LessWrong | $1,000 Bounty for Pro-BLM Policy Analysis
I want to know what Americans with no particular relevant expertise should do in order to achieve the policy goals of the Black Lives Matter movement in the United States. I'm offering a $1,000 bounty (payable directly to you or to a charity of your choice) to whoever provides... |
6814af06-ad9e-46a7-bf86-19b18461b87c | trentmkelly/LessWrong-43k | LessWrong | Delta variant: we should probably be re-masking
Like you, I thought I saw the light at the end of the tunnel. It first appeared when vaccines started rolling out a few months back, and got brighter when my family got their vaccine cards signed. It got still brighter as I signed up for my first dose (#ModernaManiacs), ... |
9ac2c24a-6577-4782-a082-ca06bd28c036 | trentmkelly/LessWrong-43k | LessWrong | In praise of contributing examples, analogies and lingo
Examples, analogies and lingo all make posts easier to understand. So then, I want to encourage people to write more comments like:
* X is an example of what you're describing.
* Y might be a good analogy for this.
* Z could be a good term for that.
I see co... |
a5d735ea-96b8-469f-98c8-3a86c59adc6e | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Stampy's AI Safety Info - New Distillations #4 [July 2023]
Hey! This is another update from the distillers at [AI Safety Info.](https://aisafety.info/)
Here are a couple of the answers that... |
c2dc756c-0cef-4067-8382-54f1e4e51773 | trentmkelly/LessWrong-43k | LessWrong | Meetup : San Francisco Meetup: Fun and Games
Discussion article for the meetup : San Francisco Meetup: Fun and Games
WHEN: 17 August 2015 06:15:00PM (-0700)
WHERE: 1390 Market St. San Francisco
We'll be meeting to hang out and play board games at my apartment (our normal place doesn't reliably have tables). Please ... |
267830ee-8634-49db-9613-15b0e4772d76 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Brussels monthly meetup
Discussion article for the meetup : Brussels monthly meetup
WHEN: 14 September 2013 01:00:00PM (+0200)
WHERE: Rue des Alexiens 55 1000 Bruxelles
Brussels meetups now happen every month on the second Saturday of the month (even when we forget to announce them in time...).
We are once... |
3d7684f8-7e25-4f15-b6f1-84c4d4049790 | trentmkelly/LessWrong-43k | LessWrong | Indifference and compensatory rewards
Crossposted at the Intelligent Agents Forum
It's occurred to me that there is a framework where we can see all "indifference" results as corrective rewards, both for the utility function change indifference and for the policy change indifference.
Imagine that the agent has rewar... |
82ccef7d-552c-4274-a8a3-330f77e3e1d0 | trentmkelly/LessWrong-43k | LessWrong | Community norms poll (2 mins)
A poll about what community norms should be in light of the nonlinear saga. Trying to be forward-looking rather than going back over events. Votes are anonymous (though given an alphanumeric string so all your votes are connected)
Consider in regard to what you want rationalist norms to ... |
7d80f396-1f99-4744-8619-f3e4bd2513e3 | trentmkelly/LessWrong-43k | LessWrong | A taxonomy of non-schemer models (Section 1.2 of “Scheming AIs”)
This is Section 1.2 of my report “Scheming AIs: Will AIs fake alignment during training in order to get power?”. There’s also a summary of the full report here (audio here). The summary covers most of the main points and technical terms, and I’m hoping t... |
0c3d1036-5834-48ea-bc91-321d696ce72f | trentmkelly/LessWrong-43k | LessWrong | Publication decisions for large language models, and their impacts
This post is one part of the sequence Understanding the diffusion of large language models. As context for this post, I strongly recommend reading at least the 5-minute summary of the sequence.
In this post, I:
1. Overview the different forms of in... |
08d14636-fa21-4750-b98d-610ba390588d | StampyAI/alignment-research-dataset/lesswrong | LessWrong | AI Safety Hub Serbia Soft Launch
TLDR; We got three-month funding from a generous individual funder to launch an AI Safety office in Serbia. We are giving free office space (and, if funding later permits, housing) to AI Safety researchers who are looking for a place to work from. Priority for citizens of countries lik... |
1f84902e-08e4-4bdf-97bc-97e49dd5cce9 | trentmkelly/LessWrong-43k | LessWrong | Anthropically Blind: the anthropic shadow is reflectively inconsistent
For the purposes of this post, the anthropic shadow is the type of inference found in How Many LHC Failures Is Too Many?.
> "Anthropic principle! If the LHC had worked, it would have produced a black hole or strangelet or vacuum failure, and we wo... |
16801386-c4a0-4805-8004-18a69574b5c0 | trentmkelly/LessWrong-43k | LessWrong | UAP and Global Catastrophic Risks
DoD has confirmed yesterday the authenticity of UAP videos and Yudkowsky made a bet: “So apparently there’s some kind of thing right now about supposed aliens. I haven’t looked into it at all, but I am happy to blindly bet against anything to do with visible intelligent aliens at 100:... |
789c5655-ea56-47fd-87cb-9572d6ca6a25 | trentmkelly/LessWrong-43k | LessWrong | Comment section from 05/19/2019
I moved the big meta-level comment thread from "Yes Requires the Possibility of No" over to here, since it seemed mostly unrelated to that top-level post. This not being on frontpage also makes it easier for people to just directly discuss the moderation and meta-level norms. |
b934aa06-92e0-43df-af6a-aea561cf2817 | trentmkelly/LessWrong-43k | LessWrong | [link] Anger as antidote to Confirmation Bias
Abstract:
> The current research explores the effect of anger on hypothesis confirmation — the propensity to seek information that confirms rather than disconfirms one’s opinion. We argue that the moving against action tendency associated with anger leads angry individua... |
bf44e41b-6fed-48fe-b006-43f7bef0e5bb | trentmkelly/LessWrong-43k | LessWrong | Is it true that most studies in Kahneman's 'Thinking fast and slow' are not replicable and the book is largely not supported by evidence?
So I stumbled on this article.
https://replicationindex.com/2020/12/30/a-meta-scientific-perspective-on-thinking-fast-and-slow
'So, replicability [of all studies in this book] i... |
0a49e18e-14fb-4605-a28e-1604671f7ed3 | trentmkelly/LessWrong-43k | LessWrong | Research agenda for AI safety and a better civilization
This agenda is various ideas that occurred to me throughout the years of reading up on AI progress and AI safety. One big idea is that AI is simultaneously software, applied mathematics as well as philosophy and ideology. AI being philosophy and ideology means th... |
b40fc9ea-01c0-43a7-9b2b-496c85d41a2c | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "(Part 6 of 8 in "Three Worlds Collide")
Today was the day.
The streets of ancient Earth were crowded to overbursting with people looking up at the sky, faces crowded up against windows.
Waiting for their sorrows to end.
Akon was looking down at their faces, from the balcony of a room in a well-guarded hotel. There wer... |
8d1a653a-b4d0-41a2-9260-7a4159cdfe2f | trentmkelly/LessWrong-43k | LessWrong | My Most Likely Reason to Die Young is AI X-Risk
Introduction
One of my pet peeves is people calling AI Safety a longtermist cause area. It is entirely reasonable to consider AI Safety a high priority cause area without putting significant value on future generations, let alone buying things like total utilitarianism o... |
ab642f8e-73be-490d-9e6e-e232f21b7161 | trentmkelly/LessWrong-43k | LessWrong | Should the front page include the phrase "latest insights"?
The front page claims
<blockquote>We use the latest insights from cognitive science, social psychology,probability theory, and decision theory to improve our understanding of how the world works and what we can do to achieve our goals.</blockquote>
I fin... |
0c946f1c-cc61-491d-892f-3b4569b7a85c | trentmkelly/LessWrong-43k | LessWrong | OpenAI's Preparedness Framework: Praise & Recommendations
In this post, I summarize some of my thoughts on OpenAI's recent preparedness framework. The post focuses on my opinions/analysis of the framework, and it doesn't have much summary (for readers unfamiliar with the framework, I recommend Zvi’s post).
As a high... |
cfc04488-d084-4135-8510-82b88a8ac85b | StampyAI/alignment-research-dataset/lesswrong | LessWrong | I Would Have Solved Alignment, But I Was Worried That Would Advance Timelines
The alignment community is ostensibly a group of people concerned about AI risk. Lately, it would be more accurate to describe it as a group of people concerned about AI timelines.
AI timelines have some relation to AI risk. Slower timeline... |
847bfcce-f2c4-48f0-bd15-3c87d944c336 | trentmkelly/LessWrong-43k | LessWrong | Group Rationality Diary, January 16-31
This is the public group instrumental rationality diary for January 16-31.
> It's a place to record and chat about it if you have done, or are actively doing, things like:
>
> * Established a useful new habit
> * Obtained new evidence that made you change your mind about so... |
78f1a66f-8afb-4e5a-a1f7-5f59bb1968c9 | trentmkelly/LessWrong-43k | LessWrong | [Linkpost] The AGI Show podcast
This is a linkpost for www.theagishow.com. Also on YouTube at https://www.youtube.com/@theagishow.
Hey everyone, my name is Soroush and I work full-time on AGI alignment.
As part of my work in alignment, I've started a podcast, The AGI Show, to help discuss and promote the best and mo... |
aa101ec8-cae6-4570-8a87-68d995f9e75f | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Is there any policy for a fair treatment of AIs whose friendliness is in doubt?
A game theory question.
If there ever exists a roughly-human-level agenty AI who could grow to overpower humans but who humans have an opportunity to stop because takeoff is slow enough. Assume the AI could coexist with humanity but fears... |
45fc7e1d-d2dd-468e-8130-9a394f36c0da | trentmkelly/LessWrong-43k | LessWrong | Don't Count Your Chickens...
A blog post by Derek Sivers links to evidence that stating one's goals makes one less likely to accomplish them.
Excerpt:
> Announcing your plans to others satisfies your self-identity just enough that you're less motivated to do the hard work needed.
Link: Shut up! Announcing your plan... |
9139c980-7d15-4d1e-9a29-eea4906b80c1 | trentmkelly/LessWrong-43k | LessWrong | A City Within a City
My local gym is surrounded on all four sides by tent cities. At first I was nervous to walk there after work, but soon I got used to the tents and their residents.
When I’m passing through on my way to the gym, I feel like I’m traveling from Rome to the Vatican—a city within a city.
They have ar... |
afb167b4-fcef-41be-9592-f43a19c95bec | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "Many, many, many are the flaws in human reasoning which lead us to overestimate how well our beloved theory explains the facts. The phlogiston theory of chemistry could explain just about anything, so long as it didn’t have to predict it in advance. And the more phenomena you use your favored theory to explain, the tr... |
f9783511-0860-45a3-8414-c7163cdf1a30 | StampyAI/alignment-research-dataset/blogs | Blogs | Why MIRI?
I wrote a short profile of MIRI for a forthcoming book on [effective altruism](http://lesswrong.com/lw/hx4/four_focus_areas_of_effective_altruism/). It leaves out [many important details](http://ukcatalogue.oup.com/product/9780199678112.do), but hits many of the key points pretty succinctly:
> The [Machin... |
46f27cba-0d60-4d9c-a99f-c54832fb305f | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Series of absurd upgrades in nature's great search
So you want to find that special thing that replicates best and lasts longest? Just vibrate a bunch of molecules a long time!
You might reasonably assume that molecules wouldn't practically ever randomly assemble themselves into anything worth looking at. It happens... |
8e2456cc-59a8-4321-99e0-ea5e988a64a1 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Slowing down AI progress is an underexplored alignment strategy
The emotional burden of watching the world end
==============================================
My current beliefs about AI timelines have made my life significantly worse. I find thoughts about the ever-shrinking timeline to AGI invading nearly every aspe... |
8b668070-bc48-4a70-b474-cd26f6c652c5 | trentmkelly/LessWrong-43k | LessWrong | Maybe you want to maximise paperclips too
As most LWers will know, Clippy the Paperclip Maximiser is a superintelligence who wants to tile the universe with paperclips. The LessWrong wiki entry for Paperclip Maximizer says that:
> The goal of maximizing paperclips is chosen for illustrative purposes because it is ver... |
012257b2-d1bf-4dc2-81f2-d56289087594 | trentmkelly/LessWrong-43k | LessWrong | Extreme Rationality: It Could Be Great
Reply to: Extreme Rationality: It's Not That Great
I considered making this into a comment on Yvain's last post, but I'd like to redirect the discussion slightly. Yvain's warning is important, but we're left with the question of how to turn the current state of the art in ration... |
8ff10fe4-7647-4b5e-963a-550e15cf4858 | trentmkelly/LessWrong-43k | LessWrong | Information throughput of biological humans and frontier LLMs
Biological humans appear, across many domains, to have have an information throughput of at most about 50 bits per second. Naively multiplying this by the number of humans gives an upper bound of about 500 gigabits per second when considering the informatio... |
eac26797-d46b-4c5d-8a27-d49cb75ce7cc | trentmkelly/LessWrong-43k | LessWrong | The Importance of Mathematics (Gowers)
For the past few days I've been pondering the question of how best to respond to paulfchristiano's recent posts and comments questioning the value of mathematical research. I don't think I can do it concisely, in a single post; bridging the inferential distance may require someth... |
3b84d9f2-53f1-431a-99a0-238d6b45a99c | trentmkelly/LessWrong-43k | LessWrong | Self Service Roundup
Follow-up to: Full Service
Before moving on to other topics, I thought I would round out my thoughts on self/full service itself (for some good follow-up on the drugstores, see the comments of the previous post) about the History of Self-Service, which I found fit my model of such things quite we... |
ac8d6da9-0d6c-4884-a12a-6da0912f6b49 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | [AN #55] Regulatory markets and international standards as a means of ensuring beneficial AI
Find all Alignment Newsletter resources [here](http://rohinshah.com/alignment-newsletter/). In particular, you can [sign up](http://eepurl.com/dqMSZj), or look through this [spreadsheet](https://docs.google.com/spreadsheets/d/... |
e122845b-4051-4f3a-befb-a1359f6b61d3 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Specific versus General Principles for Constitutional AI
1 Introduction
---------------
The method of Constitutional AI (CAI) [[1](#bib.bib1)] makes it possible to train a harmless AI assistant via self-improvement, without requiring any human supervision to identify harmful outputs. It generalizes Reinforcement Le... |
2715564d-a9c3-444b-90d5-7471c301b3a2 | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "There has been considerable debate over whether development in AI will experience a discontinuity, or whether it will follow a more continuous growth curve. Given the lack of consensus and the confusing, diverse terminology, it is natural to hypothesize that much of the debate is due to simple misunderstandings. Here,... |
a807455f-471a-40a3-bb89-31571c2cb1fe | trentmkelly/LessWrong-43k | LessWrong | If You Demand Magic, Magic Won't Help
> Most witches don't believe in gods. They know that the gods exist, of course. They even deal with them occasionally. But they don't believe in them. They know them too well. It would be like believing in the postman.
> —Terry Pratchett, Witches Abroad
Once upon a t... |
459162ca-0c00-4bd0-9288-21ed6d4ab8c3 | trentmkelly/LessWrong-43k | LessWrong | A Guide to Rational Investing
Hello Less Wrong, I don't post here much but I've been involved in the Bay Area Less Wrong community for several years, where many of you know me from. The following is a white paper I wrote earlier this year for my firm, RHS Financial, a San Francisco based private wealth management prac... |
fe94364a-5b06-4cda-98cf-2e0a02dc02ed | trentmkelly/LessWrong-43k | LessWrong | Happiness Logging: One Year In
I've been logging my happiness for a year now. [1] My phone notifies me at unpredictable intervals, and I respond with some tags. For example, if it pinged me now, I would enter "6 home bed computer blog". I always have a numeric tag for my current happiness, and then additional tags for... |
32429285-95c9-42d3-9303-488aa7c06a1f | trentmkelly/LessWrong-43k | LessWrong | Enhanced Clarity to Bridge the AI Labeling Gap?
I’ve noticed a lot of conversations here about AI safety and whether an AI can have a certain “mindset” based on how it’s trained or programmed. It’s a fascinating topic, but it can also be confusing. Especially for the people with no familiarity with AI.
One idea that... |
e3dba83a-621d-4a04-a2b3-78a7be17223c | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "When I'm learning a new skill, there's a technique I often use to quickly gain the basics of the new skill without getting drowned in the plethora of resources that exist. I've found that just 3 resources that cover the skill from 3 separate viewpoints(along with either daily practice or a project) is enough to quickl... |
56899333-72d7-4f0f-8349-a322514e326e | trentmkelly/LessWrong-43k | LessWrong | Article Review: Discovering Latent Knowledge (Burns, Ye, et al)
[This is a cross-post from my blog at aizi.substack.com.]
Large Language Models are hell-demons summoned from the data dimension. You feed a giant machine approximately everything ever written, and suddenly it can have a conversation, run a virtual machi... |
c140a1c6-23f9-40eb-801c-d371dea71f1a | trentmkelly/LessWrong-43k | LessWrong | Criticism of some popular LW articles
My composition teacher in college told me that in some pottery schools, the teacher holds up your pot, examines it, comments on it, and then smashes it on the floor. They do this for your first 100 pots.
In that spirit, this post's epistemic status is SMASH THIS POT.
As an exper... |
0dbcdb02-897e-4ee9-8f71-ec867b25d356 | trentmkelly/LessWrong-43k | LessWrong | Decoherence
Previously in series: Feynman Paths
To understand the quantum process called "decoherence", we first need to look at how the special case of quantum independence can be destroyed—how the evolution of a quantum system can produce entanglement where there was formerly independence.
Quantum independence,... |
3b325f97-1cf7-4d8b-9bfb-d4e5ea1df3c4 | trentmkelly/LessWrong-43k | LessWrong | Can we “cure” cancer?
In an excellent recent essay on “big visions for biology,” Sam Rodriques writes:
> Ask most biologists about the cure for cancer, and they will tell you it doesn’t exist: cancer is many diseases that are mostly unrelated to each other, and that all have to be cured one at a time.
Are “most biol... |
004a2225-8c5b-4524-86c5-a53fb42c7719 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Help me to understand AI alignment!
As will be very clear from my post, I'm not a computer scientist. However, I am reasonably intelligent and would like to improve my understanding of AI risk.
As I understand it (please do let me know if I've got this wrong), the risk is that:
* an AGI could rapidly become many tim... |
fabdb828-f70f-49c6-b219-5ece9122a059 | trentmkelly/LessWrong-43k | LessWrong | The Road to Mazedom
Previous post: How Escape From Immoral Mazes
Sequence begins here: Moloch Hasn’t Won
The previous posts mostly took mazes as given.
As an individual, one’s ability to fight any large system is limited.
That does not mean our individual decisions do not matter. They do matter. They add up.
M... |
d1c58f84-4cec-4ec4-8f3b-22414514e8be | trentmkelly/LessWrong-43k | LessWrong | The evolution of superstitious and superstition-like behaviour
|
684e0ad6-2733-4053-a3d3-a5a225dcf2a0 | trentmkelly/LessWrong-43k | LessWrong | Meetup : West LA—The Merits of Specificity
Discussion article for the meetup : West LA—The Merits of Specificity
WHEN: 13 November 2013 07:00:00PM (-0800)
WHERE: 10850 West Pico Blvd, Los Angeles, CA 90064, USA
How to get in: Go to the Westside Tavern in the upstairs Wine Bar (all ages welcome), located inside the ... |
ee85f131-87d6-4d45-aeb9-cab91a4c600e | trentmkelly/LessWrong-43k | LessWrong | A concise definition of what it means to win
A concise definition of what it means to win[1]
Amor vincit omnia
What does it mean for AI alignment to have “gone well”? Many answers have been proposed, but here is mine. A few basic requirements:
* We don’t all die immediately, or in a few months, or a few years (the... |
05053979-8743-49a1-a58a-131a1ace1c6b | trentmkelly/LessWrong-43k | LessWrong | Introduction to Pragmatic AI Safety [Pragmatic AI Safety #1]
This is the introduction to a sequence of posts that describe our models for Pragmatic AI Safety. Thanks to Oliver Zhang, Mantas Mazeika, Scott Emmons, Neel Nanda, Cameron Berg, Michael Chen, Vael Gates, Joe Kwon, Jacob Steinhardt, Steven Basart, and Jacob H... |
9c5deaa5-8d5c-40b7-bec2-93513cd51db1 | trentmkelly/LessWrong-43k | LessWrong | Math: Textbooks and the DTP pipeline
Epistemic status: Half-certain, half-skeptical. This mental model feels a bit too compressed to me to be maximally useful.
I've spent a surprising (=non-zero) number of hours studying mathematics during my summer break, and noticed an interesting pattern in undergraduate textbooks... |
80b97a96-b491-4d10-8dba-88b5da4e0a57 | trentmkelly/LessWrong-43k | LessWrong | Narcissism vs. social signalling
The main thrust of Robin Hanson's work is that so much of human behaviour is the result of social signalling; that is, the attempt to convince others that we possess good qualities. On the other hand, The Last Psychiatrist presents narcissism as an alternative theory; that is, much of ... |
97627de0-0a98-4364-8a4f-16abe4b86391 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post925
Summary: Many proposed AGI alignment procedures involve taking a pretrained model and training it using rewards from an oversight process to get a policy . These procedures might fail when the oversight procedure is locally inadequate: that is, if the model is able to trick the oversight process into givin... |
a20cc732-c8e3-4ac6-9fc5-7027c1ca7f54 | trentmkelly/LessWrong-43k | LessWrong | A physicist's approach to Origins of Life
Note: this is cross-posted from the post here, where I had a few videos that I could not add in here - see the original post for the full experience
I spent the 6 years in my Ph.D. thinking of the broad topic of origins of life, or “abiogenesis”: how does non-living matter tu... |
54bd00da-e240-4665-9d9f-fc5191d68dfe | trentmkelly/LessWrong-43k | LessWrong | Upcoming Protest for AI Safety
The PauseAI movement is planning protests for Friday, February 28th in several American metropolitan areas.
I understand some people's reservations about protests causing the public to develop negative associations with AI safety. In response, I invite them to join us as a way of direct... |
5c7e1ad9-033c-4e21-b16e-32643c9ec796 | trentmkelly/LessWrong-43k | LessWrong | Does there exist an AGI-level parameter setting for modern DRL architectures?
Suppose the architecture includes memory (in the form of a recurrent state) and will act as the policy network for an observation-based RL agent. Evaluating the agent from a reasonable initial state, would you guess that there exists a model... |
e609cc7d-bb26-4f3f-9310-e6cc24eb855a | StampyAI/alignment-research-dataset/blogs | Blogs | White House submissions and report on AI safety
In May, the White House Office of Science and Technology Policy (OSTP) [announced](https://www.whitehouse.gov/blog/2016/05/03/preparing-future-artificial-intelligence) “a new series of workshops and an interagency working group to learn more about the benefits and risks ... |
26f25cba-4d50-4440-80f4-a50ec68a82f3 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Outperforming the human Atari benchmark
I haven't paid much attention to Atari in a long time, and so would appreciate takes from anyone who follows this more closely. My take:
A single architecture that can handle both the games that require significant exploration, and the games that require long-term credit assign... |
2a8f0c0f-4d9b-4e59-be11-0151079d2018 | trentmkelly/LessWrong-43k | LessWrong | List #3:
Why not to assume on prior that AGI-alignment workarounds are available
A friend in technical AI Safety shared a list of cruxes for their next career step.
One premise implicit in their writing was that a long-term safe AGI can be build in the first place (or that we should resolve to build safe AGI, ... |
3c1c0ac0-9e47-4f49-aa16-c270a2a9e830 | StampyAI/alignment-research-dataset/arbital | Arbital | Join and meet: Examples
A union of sets and the least common multiple of a set of natural numbers can both be viewed as joins. In addition, joins can be useful to the designers of statically typed programming languages.
%%%knows-requisite([https://arbital.com/p/3v1](https://arbital.com/p/3v1)):
The real numbers
-----... |
5efc6dc7-cccf-4cab-bb12-67a9cf5c1e3a | trentmkelly/LessWrong-43k | LessWrong | Understanding Shapley Values with Venn Diagrams
A huge thanks to @Agustín Covarrubias 🔸 for his feedback and support on the following article:
----------------------------------------
Shapley values are an extremely popular tool in both economics and explainable AI.
In this article, we use the concept of “synergy... |
05c2f40d-063a-44df-ba72-a8974736607c | trentmkelly/LessWrong-43k | LessWrong | Covid 11/12: The Winds of Winter
Spoiler Alert below the fold, can’t be helped.
We have perhaps the best possible news. Pfizer’s vaccine is over 90% effective, and they are going to ask for emergency use authorization in a few weeks. Woo-hoo!
We also have very bad news. It was clear the numbers were going to get wor... |
8b6e0a7c-fcb5-422d-b5b3-22f442e77624 | trentmkelly/LessWrong-43k | LessWrong | Demanding and Designing Aligned Cognitive Architectures
This post is to announce my new paper Demanding and Designing Aligned Cognitive Architectures, which I recently presented in the PERLS workshop (Political Economy of Reinforcement Learning)] at NeurIPS 2021.
In this post, I will give a brief overview of the pape... |
02c04580-e096-44f6-a464-97adcea1e480 | trentmkelly/LessWrong-43k | LessWrong | How to train your transformer
In this blogpost I describe a transformer project, how I proceeded, the results, the pitfalls and more. At the end of the post I will repeat all salient points in a TL;DR, so if you are short of time, just skip ahead.
I originally published this on a company blog. The blog has been dow... |
62705df7-75d2-4e6a-a2c3-cd2935112502 | StampyAI/alignment-research-dataset/special_docs | Other | Social choice ethics in artificial intelligence
ORIGINAL ARTICLE
Social choice ethics in artificial intelligence
Seth D. Baum1
Received: 17 July 2016 / Accepted: 16 September 2017
/C211Springer-Verlag London Ltd. 2017
Abstract A major approach to the ethics of artificial
intelligence (AI) is to use social choice, in whi... |
c580ab7b-70a1-4f42-8b9e-a572458fe63a | trentmkelly/LessWrong-43k | LessWrong | Outcomes of the Geopolitical Singularity
We will soon enter an unstable state where the balance of military and political power will shift significantly because of advanced AI.
As different nations gain access to new advanced technologies, nations that have a relative lead will amass huge power over those that are le... |
8e8fc106-9bee-4579-9ddb-41c5963b5f44 | trentmkelly/LessWrong-43k | LessWrong | Predictions for GPT-N
Regarding GPT-3, there is some discussion whether growing the model would transform it into an Oracle AI. I looked into the actual benchmark results (Appendix H in the paper) to see if we can predict something useful from the actual measurements.
Method: The OpenAI team ran a suite of 63 differe... |
8b335fd7-e874-44ab-8895-33c39386a458 | StampyAI/alignment-research-dataset/blogs | Blogs | Applications are open for the MIRI Summer Fellows Program!
CFAR and MIRI are running our fifth annual MIRI Summer Fellows Program (MSFP) in the San Francisco Bay Area from August 9 to August 24, 2019.
MSFP is an extended retreat for mathematicians and programmers with a serious interest in making technical progress o... |
1f7718e1-736e-4b16-a21e-ff2eb04c9de4 | trentmkelly/LessWrong-43k | LessWrong | Progress links and tweets, 2023-02-15
The Progress Forum
* AMA: Matt Clancy, Open Philanthropy
* The Rise of Steel - Part I, by Brian Potter
Opportunities
* Sloan Foundation offering $75k–250k grants for history of science, technology, economics, and social science (via @SloanFoundation and @epistemographer)
* O... |
24e2d9e8-e3bd-4696-ad7f-42e9dc2b95b9 | trentmkelly/LessWrong-43k | LessWrong | Synthetic Restrictions
Introduction
As a child, I have read many stories, delving into everything from geography to Spanish literature, crime novels, and global history books. Each day, before or after school, I'd go to the local library, eagerly choosing the books that would be my companions for the week.
A few year... |
2de0ca9e-53af-4102-8725-4d3d80aecc5b | StampyAI/alignment-research-dataset/lesswrong | LessWrong | HCH and Adversarial Questions
*This is a paper I wrote as part of a PhD program in philosophy, in trying to learn more about and pivot towards alignment research. In it, I mostly aimed to build up and distill my knowledge of IDA.*
*Special thanks to Daniel Kokotajlo for his mentorship on this, and to Michael Brownst... |
819d1865-fa70-488f-a4b4-e42e0143ae0e | trentmkelly/LessWrong-43k | LessWrong | Meetup Interest Probe: Urbana-Champaign, Illinois
I'm going to be starting graduate school in Urbana-Champaign this fall, and want to know if there are LessWrongers there (or like me, going to be there) who want to hold some meetups. Google shows me that someone else tried asking the same thing last year and met with ... |
026ac1a3-f82c-4347-93f6-8ae0c2663d37 | trentmkelly/LessWrong-43k | LessWrong | Caelum est Conterrens: I frankly don't see how this is a horror story
So Eliezer said in his March 1st HPMOR progress report:
> I recommend the recursive fanfic “Friendship is Optimal: Caelum est Conterrens” (Heaven Is Terrifying). This is the first and only effective horror novel I have ever read, since unlike Love... |
4572c7b7-dfcb-4e21-8e70-fb392231f825 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Where I agree and disagree with Eliezer
(*Partially in response to*[*AGI Ruin: A list of Lethalities*](https://www.lesswrong.com/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a-list-of-lethalities). *Written in the same rambling style. Not exhaustive.*)
### Agreements
1. Powerful AI systems have a good chance of deliberately and... |
9db43633-4b8e-4f11-915c-2191d040a1a3 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Automatically finding feature vectors in the OV circuits of Transformers without using probing
Let's say that you're working to understand how a given large language model computes its output on a certain task. You might be curious about the following questions:
* What are some linear feature vectors that are respons... |
b45b2920-b56a-479c-a914-48d13f2b0bbf | trentmkelly/LessWrong-43k | LessWrong | Slowing AI: Foundations
Variables
"Slowing AI" is not monolithic. Many interventions have relevant side effects, positive or negative. There are many considerations, desiderata, and goals adjacent to slowing AI, and different interventions realize them to different degrees. This section is a list and discussion of var... |
ff758514-f335-4edb-904d-f7f955a338da | trentmkelly/LessWrong-43k | LessWrong | Philosophy: A Diseased Discipline
Part of the sequence: Rationality and Philosophy
Eliezer's anti-philosophy post Against Modal Logics was pretty controversial, while my recent pro-philosophy (by LW standards) post and my list of useful mainstream philosophy contributions were massively up-voted. This suggests a sign... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.